pxt CLI ships with the pixeltable package. It covers two surfaces:
- Catalog operations — inspect, query, and manage tables, views, and directories. Backed by a long-lived local daemon so each command takes ~40 ms after the first invocation.
-
Service deployment — run the services defined in an application file with
pxt service. Requires theserveextra (which pulls infastapi[standard]anduvicorn):
pxt auto-spawns a daemon bound to 127.0.0.1:22089. The daemon survives across shells and stays warm for subsequent commands. Override the port with PXT_PORT.
Command structure
pxt <command> --help for per-subcommand flags and examples.
Universal flags
These flags work the same way across the catalog commands that support them and are not repeated in the per-command tables below.Working directory
pxt cd sets a working directory that is prepended to relative paths in later commands, and pxt pwd prints it — the catalog analogue of a shell’s cd/pwd.
/ (e.g. pxt ls /other_dir) resolves from the catalog root, and a pxt://org:db/... URI addresses a hosted catalog. . and .. work in any path and resolve against the working directory; .. at the catalog root keeps the root.
The working directory is scoped to the invoking terminal, not the daemon globally — it is keyed by the shell’s session, sent with every command. So separate terminals have independent working directories, and it does not leak into subprocesses or agents you launch: a spawned process runs under its own session with no working directory, so its pxt commands resolve relative paths from the catalog root regardless of what you set interactively.
Because of that isolation, scripts and agents should address the catalog with absolute paths (/... or pxt://...) and ignore the working directory; it is a convenience for interactive terminal use.
Quick reference
Inspection commands
pxt ls
List entries under a directory.
pxt ls -l:
c = has at least one computed column, i = has at least one index.
pxt describe
Show a table’s schema and metadata. The plain form is human-readable; --json returns the full get_metadata() dict.
pxt columns / pxt computed
List columns for one or more tables. pxt computed is shorthand for pxt columns --computed. The path argument may be a single table or a directory; a directory path lists columns for every table beneath it, recursively. A directory path may be a local path or a hosted URI (pxt://org:db/...). With no path, every table in the in-process catalog is listed.
pxt idxs
List indexes. Shows both B-tree and embedding indexes by default; the --embedding flag restricts to embedding indexes. Like pxt columns, the path may be a single table or a directory (walked recursively), a hosted database root (pxt://org:db), or omitted for the whole in-process catalog.
pxt history
Show a table’s version timeline.
pxt status
Daemon and runtime state: pxt version, daemon PID, configured paths, total tables, total errors.
pxt config
Every documented configuration setting with its current value and source (env, file, or unset). Credentials show <redacted> when set; the source column reveals presence even when the value is masked.
Query commands
pxt rows
Show the first N rows of a table. Unstored computed columns are skipped by default (selecting one forces evaluation, which can invoke LLMs or expensive compute); pass them explicitly via --cols to include them.
pxt get
Look up a single row by primary key. An Int, Float, or UUID PK column parses its value and rejects one that does not parse. A String PK column keeps the value as typed, so 007 stays 007. Lookup by a PK of any other type, such as Date or Timestamp, is not supported yet and finds no row. The table must declare a primary key. Unstored computed columns are skipped unless requested explicitly via --cols (consistent with rows).
pxt count
pxt errors
List rows where a stored computed column failed. The table must have a primary key (so each failing row can be identified).
Mutation commands
Every mutation accepts the universal-n/--dry-run and --json flags. The destructive ones (drop, drop-dir, recompute, revert) also prompt [y/N] with a TTY and accept -f/--force to skip the prompt; in non-interactive contexts they refuse to proceed without -f. rename and mv don’t prompt: renaming or moving a catalog entry is reversible and doesn’t lose data.
pxt drop
Drop a table or view. Use pxt drop-dir for directories.
pxt drop-dir
Remove a directory. Use pxt drop for tables/views.
pxt rename
Rename in place; the parent directory is preserved. <new_name> must be a single name component (no / or .). Takes only universal flags.
pxt mv
Move a table/view/dir under a different directory; the entry’s own name is preserved. <new_dir> can be '' or / for the root directory. Takes only universal flags.
pxt recompute
Recompute one or more computed columns of a table. Mirrors Table.recompute_columns(), without its where
predicate.
pxt revert
Undo recent ops on a table. Each revert undoes one op; --steps repeats.
Project layout
pxt schema and pxt service read a Python file, and the tables they create refer back to the UDFs that file calls. A reference is a module path, so the file has to belong to a project. The project root is the directory with the project configuration, and every local module path is relative to it.
pxt init writes that configuration in the current directory:
pyproject.toml, the same entry is appended there as [[tool.pixeltable.database]] instead, and that section marks the root.
Every directory from the root down to a file becomes one component of that file’s module path, so each of those directory names has to be a Python identifier:
recipe.py and functions.py side by side use from functions import .... Once an application moves into a subdirectory, that subdirectory joins the path: from ad_gen.functions import ....
A recorded path is how a later process — the daemon, a serving worker, a hosted pod — finds the UDF again, so a command given a file outside any project root is refused. pxt schema check and pxt service check validate a file on its own — it imports without touching the catalog, it declares what the verb needs, and its columns refer to UDFs by paths another process can resolve:
Schema management
The commands above act on one object at a time.pxt schema works differently: you describe the tables you want in a Python file, and the CLI reconciles a catalog to that description. Provisioning an empty target and evolving an existing one are the same command, so there is no separate first-time step.
FILE is a path to a Python file. The last argument is a catalog: a local directory such as my_app, or a pxt:// URI. update creates it if it doesn’t exist.
The file
The file defines one or more models on apxt.model_base(). Each model becomes one table, named by name=. pxt schema example writes a file covering every construct the schema DSL supports, so you never have to start from a blank page and never have to look up a construct:
pxt schema example --brief writes the minimal version instead:
name: type) declares a stored column; an assignment (name = expr) declares a computed column. A model with base= becomes a view of that query’s base model.
The daemon imports the file, so it must be readable there. Its own directory is added to sys.path, so it can import modules sitting next to it.
Reviewing and updating
diff prints one line per table, then one per operation:
pxt schema update
update creates missing tables and migrates existing ones, adding and dropping columns and indexes. It takes the same flags as the other mutations, plus one of its own:
--allow-destructive is absent, update does not change the catalog and exits 3.
Some differences cannot be migrated in place: a table declared where a view exists, a changed iterator, or a column whose type or properties changed. Those are reported as UNSUPPORTED, update changes nothing, and it exits 1. Adjust the application file or the table by hand.
Exit codes
The schema commands report their outcome in the exit status, so a caller never has to parse the output:
A drift check in CI is therefore one command:
Machine-readable plans
pxt schema diff --json emits the whole plan as one object: schema_file, catalog_dir, in_agreement, tables, extras, and a summary with one count per resolution. Each entry in tables has a path, a resolution, a destructive flag, and its ops:
target is column, index, or table, and its op is add, drop, or alter. name is what it acts on — a column, an index, the differing attribute when the target is a table, or the table path for a drop — and details contains that operation’s operands, such as the type of an added column. severity is additive, destructive, or unsupported; destructive is the boolean form of the middle case. A table’s resolution is up_to_date, create, update_additive, update_destructive, or unsupported, and a create has no ops, because the create subsumes them.
These field names and values are the catalog’s own, as returned by TableModel.get_model_diff(), so a plan read from the CLI and a diff read from Python describe a change the same way.
pxt schema diff --json-schema prints the JSON Schema of that output.
update and prune return the same object with a status on every table and operation:
Every path returns the plan, including the ones that refuse before reaching the daemon, so an
--allow-destructive refusal is as machine-readable as a success: exit 3, with the offending operations marked refused and the rest skipped. prune reports its drops in a top-level ops array, each with target: "table", op: "drop", and the dropped table’s path in name.
Pruning
update only ever touches tables declared by the schema, so tables it doesn’t know about accumulate. diff lists them as extras; prune drops them. A full reconcile is update followed by prune.
Interactive shell
For agentic or scripted workloads that issue many commands in sequence,pxt shell amortizes Python startup over the session:
pxt command is available unmodified. Errors from one command don’t kill the session. Use help, exit, quit, or Ctrl-D to leave.
Output and scripting
Most catalog commands accept--json for stable, machine-readable output; Universal flags lists the commands that do not:
--json, output is column-aligned text.
The schema commands additionally report drift in their exit status (0 in sync, 2 pending, 3 refused, 1 error), so a CI gate needs no output parsing.
Serving
pxt service runs the services in an application file. An application file is any Python source file containing
table/view models and either FastAPIRouter instances (which serve routes over them) or a fastapi.FastAPI
application of your own. It requires the serve extra
(pip install 'pixeltable[serve]'), which pulls in fastapi[standard] and uvicorn.
A service is a full FastAPI application with auto-generated OpenAPI docs
at
/docs. For the API that defines the routes, see the Python serving API.Serving your own application
A file may supply its ownfastapi.FastAPI object instead of leaving Pixeltable to build one. Then the
file declares one service, named after its module, and pxt service update serves that application as it
is:
Notes.insert(...)).
All FastAPIRouter instances in the same source file are expected to be included in the FastAPI application
(via include_router()), and the application file still produces a single service:
app would never be served, and in that situation pxt service fails with an error.
The routes of an included FastAPIRouter are diffed by their declarations (i.e., they take the data types of path
parameters into account); the paths served by the application itself are diffed as path strings.
Signing in
Hosted commands need a credential.pxt login, pxt whoami, and pxt key require Pixeltable 0.7.10 or later. There are three: a pxt login session, a pxt new trial (Trial databases), and an API key. You need one of them.
pxt login prints a code and opens a browser. Approve it there, signing in or signing up, and the
session lands on this machine. It uses the OAuth device grant, so nothing
listens on a port and the browser need not be on this machine: this works over SSH.
pxt whoami
reports whether Pixeltable Cloud recognizes the credential, and exits nonzero when it does not; it
does not report whether the next command will succeed. A credential that is recognized but not
permitted to list organizations, such as a key whose grants do not cover that, still counts as
accepted: pxt whoami prints the refusal as a note and exits 0. pxt whoami --offline reports the
cached session without asking.
pxt logout forgets this machine’s cached session, or its pxt new trial (Trial databases). When the session has the sign-in service’s
session id, pxt logout also opens that service’s sign-out page for the session in your browser,
and prints the link when no browser opens. It signs out of both because a browser that is still
signed in confirms the next code without naming the account, so signing out of only one leaves you
signed in as someone you did not choose. Forgetting the session needs no network. When pxt logout
cannot work out the sign-out link, for example because the control plane cannot be reached, it
still signs this machine out and warns that the browser was not signed out. Whether the sign-out
page itself loads is something only the browser shows.
An API key takes precedence over a sign-in whenever both are present: it is the explicit
choice, and what CI runs on. Set one with
PIXELTABLE_API_KEY, or with api_key in the
[pixeltable] section of your Pixeltable config file; the environment variable wins when both are
set. Unlike a sign-in, it does not expire. pxt whoami says which one your commands will actually
send.pxt login also selects
it. A brand-new account belongs to no organization, and every hosted command needs one, so pxt login
and pxt whoami print “No organization yet” with the command that creates one, pxt org create NAME,
rather than letting the next command fail as an authorization error.
Trial databases
pxt new needs no account. It creates a trial organization with one database, main, and caches
the trial’s API key on this machine, so later commands use it without pxt login or an exported
key. It prints the database’s URI, the [[pixeltable.database]] entry for pixeltable.toml, the
next commands, and a claim link. It never prints the key, and neither does pxt new --json.
Until it is claimed, the organization has 1 database and at most 2 services, each on 1 worker with up
to 1 cpu, 2048 MB and 10 GB of disk, and 50 GB of media storage: past that, uploads stop. The server sets
its expiry, normally 48 hours after creation. Its pods reach the internet on ports 80, 443 and 5432 only, and one
network holds at most 10 unclaimed trials at once. Whoever opens the claim link signs in or signs up and
becomes the organization’s admin, so keep the link out of shared logs.
Running pxt new again prints the same trial, without the claim link, and creates nothing.
pxt whoami shows the trial and its expiry. pxt logout removes the trial’s key from this machine
and prints the claim link once more. It revokes nothing: the key works until the organization is
claimed or expires. An API key or a pxt login session outranks a trial, so with either one
configured, pxt new exits 1 and creates nothing. When the trial’s key is rejected, the
organization may have been claimed, and its person signs in with pxt login.
PIXELTABLE_SITE_URL points pxt new at a site other than https://www.pixeltable.com. It must
use https, except on localhost.
Organizations
pxt://acme:main, and it must be unique across Pixeltable. It works with a pxt login session or
with a personal API key, one created without grants; a key with grants belongs to its organization
and cannot create another.
Creating one also switches your pxt login session to it: Pixeltable Cloud takes the organization
from your session, not from the URI. If the switch fails, the organization exists anyway, and
pxt org create warns and says to run pxt login. An API key stays bound to its own organization
and outranks a pxt login session, so to work in the new one, use a key created there, or remove
the API key and then run pxt login. pxt org list shows the ones you can reach.
Keys
A key is the credential that does not expire, for the places a browser sign-in cannot reach: CI, a cron job, a deployed service. There is one command and two shapes, and the grants decide which.Keys with grants are a preview. Until they are enabled for your organization, Pixeltable Cloud
refuses
pxt key create --grant with “Keys with grants are not available yet”.
A service is addressed by its base path and name, as in
access:pxt://acme:main/services/my_dir/ingest.
manage applies to services only, so manage:pxt://org:db is refused.
access and manage are independent: a key that can call a service cannot reconfigure it, and one
that can stop it cannot read what flows through it.
The secret is shown once, when the key is created, and is never retrievable afterwards. Deleting a
key revokes it. Every member of the organization sees every key, can change any runtime key’s
grants, and can delete any key, and pxt key list shows who created each one.
Hosted targets
The last argument may be apxt://org:db URI, and the services then run in that database rather than on this
machine. The database needs this project and its tables first, so the order is:
diff, update, prune, stop, restart and list all accept a hosted target. run does not: it serves from the
calling process, so it is local by definition. Two verbs differ in meaning against a hosted database, where
a stopped service keeps its registration: stop stops it and leaves it there to be started again, while
prune forgets it.
pxt service stop accepts a service address in several forms. A bare name matches the services of the
current project, or every local service when the working directory is outside a project. my_dir/NAME
stops the service under my_dir, including one in another project, and /NAME stops the service at the
catalog root. A bare name that matches more than one service is refused, and the error lists the
qualified path of each match.
skipped with
the reason (not found or already stopped), so stopping a set of services can be repeated safely.
pxt service restart cycles a service’s pods onto the database’s current image and project, keeping the
service’s registration and endpoint, and accepts the same service addresses as stop, except that a
bare name matches every local service of that name, whatever its project. Use it after
pxt db update to move a running service onto the new project, and to pick up the current values of its
referenced secrets. Against a local target it stops the process and starts it again on the same
port, so callers keep their address.
pxt service logs reads a service’s log and accepts the same service addresses as restart. pxt db logs reads
the log of a hosted database’s pod:
--include-health is given. The log outlives the pod, so the log of a failed deploy can be read after its pod is
gone. A line appears in the log a few seconds after it is written. A service running on this machine logs to a
local file instead, and pxt service logs reports the path of that file.
pxt service verbs
Like the
schema verbs, diff reports drift in its exit status: 0 in agreement, 2 changes pending,
3 refused, 1 error. So a CI gate needs no output parsing.
pxt service diff --json-schema prints the JSON Schema of that output.
Flags
Tracing
--otel emits OpenTelemetry traces from the served application: Pixeltable’s own spans, nested under the
request spans of the FastAPI app. It needs the instrumentation package (pip install 'pixeltable[otel]'),
and the endpoint and service name come from the OTEL_* environment variables or the [otel] config
section, as they do for any Pixeltable process.
--otel restarts the service, and pxt service diff app.py my_dir --otel reports that
restart as a pending change before update performs it.
Background and foreground
update starts one background process per service, each on its own port, and records it so list and
stop can find it again. A service that crashes disappears from list with no cleanup step, because a
record is only as live as the process it refers to.
run stays in the foreground until you interrupt it and does not register the service for list or stop,
so it is a separate command, not a flag on update. Use it as a container entrypoint or a development loop.
It serves one service per process, the same as update; name a specific service as a third argument when the file
declares more than one:
Restarts
A service binds its models once, when its process starts, so a changed declaration takes effect whenpxt service update replaces
the process. diff reports which services that replacement would interrupt, and update reports them
before it proceeds. Adding a route is additive; changing or removing one stops serving a contract that
callers may be using, so it needs --allow-destructive.
Cloud
pxt db and pxt org manage cloud-hosted databases and organizations. Cloud commands need a credential: a pxt login session, a pxt new trial (Trial databases), or an API key (Signing in). To use an API key, create it in the Cloud dashboard (Get an API key) or with pxt key create, then export it as PIXELTABLE_API_KEY. The CLI reads the process environment and does not load .env on its own (Configuration).
Every cloud command except pxt logout accepts --json for machine-readable output.
Cloud configuration reference
pxt db diff and pxt db update read the entry of the target database from the project configuration (pixeltable.toml, or pyproject.toml under [tool.pixeltable]), searched for upward from the working directory.
[[pixeltable.database]] — what a project declares about one database
Defines the attributes of a specific Pixeltable database (which may be either the local database, or a cloud-hosted database), including:
- Which project files to include
- Which dependencies to include
- Compute settings
- Media destinations
- Telemetry configuration
name configures the local database. A hosted database is configured by the entry whose name is its URI.
pxt db
Manage cloud-hosted Pixeltable databases. A database is a hosted Pixeltable instance with its own compute, storage, and Python environment.
Database URIs use the form pxt://org:db. Valid states: PROVISIONING, STARTING, AVAILABLE, UPDATING, STOPPING, STOPPED, FAILED.
The URI argument is optional: when it is omitted, these commands use db_uri from the Pixeltable config file (see Configuration), so a project that sets it can run pxt db status and the other db commands with no argument.
pxt db update
update creates a hosted database and keeps it current afterwards. Put the database in a [[pixeltable.database]] entry, then run update against the URI in its name:
name: an entry for another database configures that one, and a target with no entry is an error. The first run creates the database, builds the image for its environment and uploads the project files that the pods run; every later run applies whatever has changed since. The cloud configuration reference lists everything an entry can declare.
update applies a new image if dependencies changed, then uploaded project files if sources changed, then one resize for any changed CPU, memory, disk, or worker counts so the pods restart only once. Secrets are set with pxt secret rather than from the project. The media destinations and the OTLP endpoint and protocol travel with the project files, which always include the project configuration.
pxt db list
pxt db status
pxt db start
AVAILABLE.
pxt db stop
pxt db restart
AVAILABLE; a stopped database is started instead.
pxt db diff
pxt db diff compares the hosted database to that config entry (see the cloud configuration reference): capacity, the Python image, and the project file archive —
so an edit to a source file is uploaded in seconds, and only a dependency change costs an image build. Exit status is 0 in agreement and 2 with changes pending; nothing is built, resized or set.
pxt db diff --json-schema prints the JSON Schema of that output.
pxt db build-image
pxt db update rebuilds only when that environment changed. The project files are uploaded with it, unless the database already has that exact archive. Polls until the build completes or fails; the pods restart on the new image and the new sources.
When a command reports an unsupported proxy protocol version and this Pixeltable is newer than the hosted database, run this rebuild against the database URI from the failed command. What that image installs is described in What to run after a change.
pxt db delete
[y/N] on a TTY; otherwise exits with code 3. Pass -f or --force to skip confirmation.
A database recreated under the same name is a different database: its workers are keyed by a new
internal id, so credentials for the old database no longer work. Treat a
delete followed by an update as a new deployment rather than a reset.
pxt org
pxt org list
pxt org status
pxt key
create prints the secret once. list never does.
update is a delta, not a replacement, so widening a key does not depend on resending everything it
already had. Revoking a grant the key never had is an error rather than a no-op, since a typo would otherwise
read as a completed revocation. It applies only to a key that has grants: one that acts as
you has nothing to change.
Names are unique per organization across both shapes, so pxt key delete NAME never has to choose
between two keys. Any member of the organization can delete any key.
pxt secret
list prints names, never values.
list prints one table for the org and its databases. Without a URI it covers your credential’s org; pxt://myorg:mydb narrows it to the org and that database. A database secret with the same name as an org secret is marked overrides an organization secret, and the org row stays in the table:
--json, each row is an object with key and scope, plus "overrides_org": true on a database secret that overrides an org secret. With nothing set, list prints No secrets for pxt://myorg. (or the database’s URI), and list --json prints [].
set and delete print the keys they changed as the same KEY and SCOPE rows, without NOTE; with --json, as key and scope objects. Earlier releases printed bare key names, and --json printed an array of strings.
A running database or service uses the values from its startup. Run pxt db restart for the database’s tables process and pxt service restart for each service. A database or service started afterwards reads the current value at startup.
What’s next
- Working with the Pixeltable CLI: hands-on cookbook for inspect, query, debug, and serve workflows
- HTTP Serving Guide: TOML config reference, Python
FastAPIRouterAPI, decorator routes - Configuration: API keys, storage paths, and environment settings